On successfully configuring the replication on a VM, it first does an initial full sync of the source VMDKs to the target datastore. If you already have the base VMDKs previously copied to the destination datastore, then only the changed blocks are replicated. The replication happens over the network using the Network File Copy (NFC) protocol. The changed blocks are transferred using ESXi's management VMkernel port group.

Once the initial sync is complete, the VR agent tracks the changed blocks using the vSCSI filter driver. It tracks, writes, and maintains a bitmap of the changed blocks. Every time a replica is created, the data transferred is copied to a redo logfile. This is done to make sure that the VM at the recovery site is not corrupted in the event of a network disruption. The redo log is committed to the base disk only after the changed blocks are fully copied, thereby making each replica crash consistent. When you configure the replication for the VM...